Thermoelectric Transport Properties of Fe-Enriched ZnO with High

Apr 3, 2015 - Citation data is made available by participants in Crossref's Cited-by Linking service. For a more comprehensive list of citations to th...
0 downloads 5 Views 2MB Size
Research Article www.acsami.org

Thermoelectric Transport Properties of Fe-Enriched ZnO with HighTemperature Nanostructure Refinement Xin Liang*,‡ School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, United States ABSTRACT: Thermoelectric properties of Fe-doped ZnO materials are investigated and correlated with the phase and microstructural evolution. Both a ZnO solid solution phase and a ZnFe2O4 spinel phase are observed. Analysis was made of temperature measurements of Seebeck coefficients as combined with the law of mixture to estimate the Fermi level in the constituent phases, which are further correlated with the Fe solute concentration in ZnO lattices and the overall electrical conductivity. In addition, the thermoelectric figure of merit is found to increase with the actual Fe content in ZnO lattices, due to the reduced thermal conductivity by point defect scattering of phonons and enhanced electrical transport via electron doping. The maximum achievable power factor of Fe-doped ZnO material is found to be similar to that of the ZnO− In2O3 system. Another important finding of the present work is the significant nanostructure refinement in 18 month old FeO1.5doped ZnO after high-temperature thermal treatment, leading to further reduced thermal conductivity, which is beneficial and promising for high-temperature thermoelectric performance. KEYWORDS: thermoelectrics, oxide, grain refinement, nanostructure



INTRODUCTION ZnO-based oxide system has a rich variety of doping choices on altering the microstructure as well as tuning the transport properties. Examples of dopant elements are Al,1,2 Bi,3 Co,4 In,5−9 Fe,10,11 Mn,12 Ni,13,14 Sb,15 Sn,16 and even dual doping.17−19 This research provides useful information on the doping effect on the structure and/or properties, while it lacks a systematic investigation of the thermoelectric properties, especially with combined observation on the microstructure. The Fe-doped ZnO system is one such example not yet deeply explored, especially on thermoelectric transport properties. Light doping of Fe into ZnO was found not to cause any change of microstructure.20,21 Nevertheless, there were a few works on Fe highly doped ZnO reporting the formation of oxide defect structure including inversion domain structure11 or antiphase modulated structure,10 both of which are in the length scale of 10 nm. However, few systematic efforts have been made to explore the thermoelectric properties of Fedoped ZnO; neither the thermal conductivity nor the electrical conductivity and the Seebeck coefficient have been investigated. It is of primary interest in the present work to look into the effect of phase equilibria and microstructure on thermoelectric properties of Fe-doped ZnO. In the present work, 2, 4, and 18 month-old (m/o) FeO1.5doped ZnO, with pure ZnO as reference, were synthesized to obtain different phase equilibria and investigate the thermoelectric properties. The 2 and 4 m/o FeO1.5-doped ZnO were annealed at 1250 °C for achieving the solid solution phase. The 18 m/o FeO1.5-doped ZnO was annealed at 1150 and 1350 °C © 2015 American Chemical Society

to reach different phase equilibria and microstructures, including a two-phase microstructure and a highly Feconcentrated ZnO solid solution phase. The compositions and heat treatment temperatures of samples are indicated on the ZnO−FeO1.5 binary phase diagram shown in Figure 1a, which was adapted from Degterov et al.’s work.22



EXPERIMENTAL SECTION

The oxide powders were prepared with a wet-chemistry gelcombustion combined approach. High-purity commercial nitrate powders Zn(NO3)2 (99.999%, Sigma-Aldrich, USA) and Fe(NO3)3 (99.99%, Sigma-Aldrich, USA) were dissolved in deionized and distilled water. After determination of ionic concentrations, each solution was weighed and mixed according to the iron concentration. The mixed solution was uniformly stirred and held at 80 °C. With addition of a few organic fuels (acrylamide, N,N′-methylenebis(acrylamide), 2,2′-azobisisobutyro-nitrile, and ammonium persulfate), the liquid solution was converted into the gel, which was then dried in a low-temperature oven at 120 °C for 12 h to remove the water. The dried gel was crushed into small pieces and sent into combustion furnace in which self-assisted reaction occurred at ∼600 °C. The combustion powders were then calcined at 800 °C for 2 h to remove the carbon and other residual organic chemicals. After being finely ground, the calcined powders were then sintered into solid pellets using current-assisted densification processing system (also known as spark plasma sintering) at 900 °C for 5 min. During the entire sintering process, the powders were under vacuum with a Received: December 24, 2014 Accepted: April 3, 2015 Published: April 3, 2015 7927

DOI: 10.1021/am509050a ACS Appl. Mater. Interfaces 2015, 7, 7927−7937

Research Article

ACS Applied Materials & Interfaces

Figure 1. Phase equilibrium of ZnO−FeO1.5 binary system. (a) Phase diagram adapted from Degterov et al.’s work.22 Sample compositions and annealing temperatures are indicated on the figure. (b) X-ray diffraction results of Fe-doped ZnO with pure ZnO data also presented. Characteristic peaks of each phase are labeled. compressive stress of 125 MPa. The sintered solid samples were postannealed in air furnace to achieve different phase equilibria and microstructures. As a reference, pure ZnO was sintered in the same way and further annealed in air at 1150 °C for 1 d. X-ray diffraction analysis was performed at a Bruker D8 Discover Diffractometer. The phases were analyzed and identified with 2011 ICDD PDF database.23 The cross sections of samples were ground and polished to 1 μm followed by the thermal etching at 1050 °C for 30 min for scanning electron microscopy (SEM) imaging and elemental analysis, which was performed on a Zeiss Super VP55 FEG-SEM with X-ray energy dispersive spectroscopy (EDS). Transmission electron microscopy (TEM) investigations were also made to look into the evolution of microstructure. Specimens were prepared in a conventional method of mechanical thinning followed by ion beam milling on a Fischione 1010 Dual Beam Ion-Mill machine. After the samples were further cleaned on a Fischione 110 Plasma Cleaner, the TEM observations were made on JEOL 2010 FEG-TEM. Scanning transmission electron microscopy (STEM) with X-ray EDS mapping analysis was also made in the purpose of understanding the elemental distribution of fine structures, which was conducted on a Cs-aberration corrected Zeiss Libra 200 MC STEM with twin EDS detectors. Thermal conductivity (κ) was obtained from the thermal diffusivity (α), which was measured from room temperature to 800 °C in flowing argon gas on a NETZCH Micro Flash LFA 457 instrument. Heat capacity Cp was calculated using the Kopp−Neumann rule from literature data,24 and the mass density ρ was measured using Archimedes method. The thermal conductivity with temperature was then determined from the relation

κ(T ) = α(T )· ρ ·C p(T )

2θ of 10° to 70°, while spectra in the range of 2θ of 25° to 40° are presented as they are most representative of important phase and structure features. Characteristic peaks of each phase are labeled. Analysis of the X-ray diffraction results reveals several important features, as will be described in the following. With increasing Fe concentration to 4 m/o FeO1.5, a systematic shift of ZnO (002) peak toward smaller 2θ angle was observed, which indicates increasing basal plane spacing, implying an increased incorporation of Fe ions into the ZnO lattice. A continuous broadening of ZnO (101) peaks with increasing Fe concentration was also observed, as has also been reported in literature.25 In addition, a ZnFe2O4-type of spinel phase (cubic, Fd3̅m) but deficient in Fe, one example of which is a Zn0.945Fe1.78O3.71 phase (ICDD No. 01−079−1500), was found to be present in both 2 and 4 m/o FeO1.5-doped ZnO that were annealed at 1250 °C. The 4 m/o FeO1.5-doped ZnO seems to have a smaller fraction of Zn0.945Fe1.78O3.71 spinel phase than the 2 m/o FeO1.5 sample, but it has larger concentration of Fe in ZnO lattice, as seen from the larger shift of ZnO (002), (100), and (101) peaks. The X-ray diffraction results shown in Figure 1b suggest the presence of both ZnO and ZnFe2O4 spinel phases in 18 m/o FeO1.5-doped ZnO annealed at 1150 °C, which is expected from the phase diagram shown in Figure 1a. However, it is interesting that ZnO solid solution phase possesses a lower content of Fe, as clearly seen from its smaller shift of ZnO (002) peak as compared to that of 2 or 4 m/o FeO1.5-doped ZnO. The volume fraction of spinel phase was estimated to be ∼0.18. After the 18 m/o FeO1.5 sample was further annealed at 1350 °C for 3 d, two ZnO solid solution phases were identified. One ZnO solid solution phase has the (002) peak close to the pure ZnO one, whereas the other one has a significant shift and identified as the (002) peak of Zn0.750Fe0.171O, an Fe highly enriched ZnO solid solution phase (ICSD No. 155784).25 In addition, a tiny amount of Fe-deficient ZnFe2O4 spinel phase was also observed in addition to the stoichiometric ZnFe2O4

(1)

Specimens were cut into bars of dimensions ∼2 × 2 × 8 mm for electrical conductivity and Seebeck coefficient measurements, which were made in air from room temperature to 800 °C on a ULVAC RIKO ZEM 3 M 10 unit.



RESULTS Structural Characterization. X-ray diffraction results for Fe-doped ZnO, including pure ZnO as reference, are presented in Figure 1b. The X-ray scan of all the samples was made from 7928

DOI: 10.1021/am509050a ACS Appl. Mater. Interfaces 2015, 7, 7927−7937

Research Article

ACS Applied Materials & Interfaces

Figure 2. Cross-section microstructure of 18 m/o FeO1.5-doped ZnO annealed at 1150 °C for 1 d: (a) SEM micrograph; (b) EDS mapping with green and red colored data points representing Zn and Fe, respectively.

Figure 3. TEM observations of microstructure in 18 m/o FeO1.5-doped ZnO annealed at 1150 °C for 1 d: (a) TEM micrograph of grains; (b) HRTEM lattice image of ZnO {002} basal planes.

Figure 4. SEM micrographs of cross-section microstructure of 18 m/o FeO1.5-doped ZnO annealed at 1350 °C for 3 d. (a) Fine grain structures. (b) Platelet or lamellar structures.

phase, as is consistent with the X-ray diffraction results. The grain structure is clean inside as shown in the TEM micrograph in Figure 3a. In addition, no oxide defect structures such as superlattice interfaces or antiphase boundaries were observed in our high-resolution (HR) TEM investigations; one typical lattice image of ZnO {002} basal planes is presented in Figure 3b. After a further annealing at 1350 °C for 3 d, a significant structural change in 18 m/o FeO1.5 sample was observed, as shown in the SEM observations in Figure 4. The material

spinel phase. Another feature found in this sample is the peak broadening, especially the ZnO (101) peak, indicating the reduction in grain size. The 18 m/o FeO1.5-doped ZnO, with plenty of features revealed by the X-ray diffraction, was further investigated under electron microscopy. The SEM cross-section micrograph of the sample after 1150 °C annealing is shown in Figure 2a. The typical grain size was ∼2−3 μm. The corresponding EDS elemental mapping, as presented in Figure 2b, suggests the presence for ZnO solid solution phase and ZnFe2O4 spinel 7929

DOI: 10.1021/am509050a ACS Appl. Mater. Interfaces 2015, 7, 7927−7937

Research Article

ACS Applied Materials & Interfaces

Figure 5. Cross-section microstructure of 18 m/o FeO1.5-doped ZnO annealed at 1350 °C for 3 d. (a) SEM image. (b) EDS mapping with green and red colored data points representing Zn and Fe, respectively.

Figure 6. Bright field TEM micrographs of typical microstructures in 18 m/o FeO1.5-doped ZnO annealed at 1350 °C for 3 d: (a) lamellar and (b) fine grain structures.

evolves into two types of structures. One is fine grain structure with grain size reduced to ∼200 nm, as shown in Figure 4a; the other is platelet or lamellar-type structure embedded in a large grain as shown in Figure 4b. The SEM and the corresponding EDS elemental map in Figure 5a,b, which captures both types of structures, suggest that these fine grains and platelets are Ferich phases and therefore are the ZnFe2O4 spinel phase that was identified by X-ray diffraction analysis. Performing an imaging analysis of the EDS mapping gives a volume fraction of 0.17 for ZnFe2O4 spinel phase. The phase compositions and fractions in 18 m/o FeO1.5 sample remain virtually unchanged between the two thermal treatments, whereas the microstructure and morphology have undergone a significant evolution. TEM investigations observed on ZnO {002} planes revealed more details of the lamellar or platelet structures in the 18 m/o FeO1.5 sample annealed at 1350 °C, as show in Figure 6a. Although there are a few reports of the presence of superlattice structure (SL) or inverse domain boundaries (IDBs) in Fedoped ZnO where Fe−O layers or IDBs stack along c-axis of the ZnO,10,11,26−29 no such structures or related defect contrast were observed in this material even in our extensive HRTEM observations of ZnO {002} planes. In addition, it is noted that the average platelet spacing is ∼200 nm, which is much larger than the average superlattice or IDB spacing, which is only a few nm. Furthermore, the platelet is much thicker than the superlattice interface or IDBs, which are only one to two atomic

layers thick. Fine grain structures of a few hundred nanometers in size, as seen in the SEM micrograph in Figure 4a, were also observed under TEM, as shown in Figure 6b. A further STEM EDS elemental map was made for ZnO {002} planes, as shown in Figure 7, suggesting that the thin lamellar structures are enriched in Fe and therefore in the ZnFe2O4 spinel phase, while the matrix is a ZnO solid solution phase. A line scan profile is presented in Figure 7c, which provides semiquantitative information on the Fe and Zn distribution in a direction perpendicular to the platelet structure. Since the scan was taken from the edge of the specimen toward inside, as indicated by the arrow on the image in Figure 7a, the specimen thickness increases away from the edge, leading to the increasing energy dispersive analysis of Xrays signal intensity for both zinc and iron. The counts ratio of Fe to Zn across these thin Fe-rich platelets is ∼2, which basically corresponds to the Fe/Zn ratio in ZnFe2O4 spinel compound. In addition, it is found that ZnFe2O4 platelets grow out of ZnO obeying the previously reported crystallographic orientation relationship of ZnFe2O4(111)[110]//ZnO(0001)[112̅0].30 The STEM EDS elemental map was also made on the fine grain structures as shown in Figure 8. Elemental distribution results clearly show that these fine grains, which have the grain size of ∼200 nm, are ZnFe2O4 spinel phase embedded in the ZnO solid solution matrix. 7930

DOI: 10.1021/am509050a ACS Appl. Mater. Interfaces 2015, 7, 7927−7937

Research Article

ACS Applied Materials & Interfaces

Figure 7. STEM image and elemental mapping of lamellar structure in 18 m/o FeO1.5-doped ZnO annealed at 1350 °C for 3 d. (a) STEM dark field image with an arrow showing the line scan direction; (b) STEM EDS map with Zn and Fe represented by green and red colored data points, respectively; (c) STEM EDS line scan across several platelets showing the Zn and Fe profile. Clearly, the thin lamellar structures are highly enriched in Fe, and the ratio of Fe to Zn is ∼2, in correspondence with ZnFe2O4 spinel phase.

Figure 8. Aberration-corrected STEM image and elemental mapping of fine grain structures in 18 m/o FeO1.5-doped ZnO annealed at 1350 °C for 3 d. (a) STEM dark field image; (b) STEM EDS map with Zn and Fe represented by green and red colored data points, respectively.

Thermal and Electrical Transport Properties. Measured thermal conductivities of Fe-doped ZnO with temperature up to 800 °C are presented in Figure 9a. Doping ZnO with Fe results in both a significant drop in thermal conductivity as well as a reduced temperature dependence. It is also interesting to see that the 18 m/o Fe O1.5 sample annealed at 1150 °C has higher thermal conductivity than the 2 or 4 m/o Fe O1.5 samples, which are less Fe-doped; however, a further annealing at 1350 °C reduces the thermal conductivity by a factor of more than 2.

Electrical conductivity and Seebeck coefficient measurement results with temperature are presented in Figure 9b,c, respectively. Electrical conductivity of ZnO is significantly improved by Fe doping. In addition, for all the Fe-doped ZnO, electrical conductivity increases with temperature. The 18 m/o FeO1.5 sample annealed at 1350 °C has the highest electrical conductivity, followed by the 4 m/o and 2 m/o FeO1.5 samples annealed at 1250 °C and the 18 m/o FeO1.5 sample annealed at 1150 °C. For all the samples, Seebeck coefficient has exactly the opposite trend as that of electrical conductivity, namely, the 7931

DOI: 10.1021/am509050a ACS Appl. Mater. Interfaces 2015, 7, 7927−7937

Research Article

ACS Applied Materials & Interfaces

Figure 9. Thermal conductivity (a), electrical conductivity (b), Seebeck coefficient (c), and figure of merit ZT (d) of Fe-doped ZnO samples measured from room temperature to 800 °C. The solid lines are for visual guide.

of Fe ions in ZnO lattices, as seen in the phase diagram in Figure 1a. Both stoichiometric and Fe-deficient spinel phase formed in this sample, but the Fe-deficient spinel phase only has a small fraction. It is clear from the X-ray diffraction results that both low Fe concentrated ZnO and Fe-deficient spinel are minor phases. Considering 2 and 4 m/o FeO1.5 sample (annealed at 1250 °C) consists of both less Fe concentrated ZnO and Fe-deficient spinel phase, it is highly possible that the less Fe concentrated ZnO solid solution phase in the 18 m/o FeO1.5 sample (1350 °C) was formed during cooling when the Fe solubility in ZnO lattice decreases and that equilibrium favors ZnO with less Fe content. The formation of a small amount of Fe-deficient spinel phase is also expected to occur during cooling and follow the similar scenario as of 2 or 4 m/o FeO1.5 sample, which was discussed previously. Detailed discussions on the phase and microstructure evolution in the 18 m/o FeO1.5 sample will be given in next section. To evaluate the lattice parameter with the Fe concentration of the ZnO solid solution phase, the ZnO (002) basal plane spacing or (c/2), where c is the unit cell lattice parameter, was obtained and calculated from the peak fitting of each X-ray diffraction data and plotted against m/o FeO1.5 in Figure 10a. Referring to the phase diagram in Figure 1a, the 2 and 4 m/o FeO1.5 compositions equilibrated at 1250 °C fall in the ZnO solid solution regime. Accordingly, it is reasonable to assume the lattice parameter of these two samples with that of pure ZnO follow the Vegard’s law, which suggests a linear dependence of lattice parameter with solute concentration.31 Together with the undoped ZnO, a straight line can be fitted through the data representing the Vegard’s law prediction, as shown by the dashed green line on the figure. Imposing the

sample that demonstrates high electrical conductivity has small Seebeck coefficient. These findings suggest that all these samples are of typical free electron semiconductors.



DISCUSSIONS Phase Equilibria and ZnO Solid Solution. At relatively low Fe concentration, such as 2 and 4 m/o FeO1.5, the phase diagram shown in Figure 1a predicts the single ZnO phase. However, in addition to the ZnO solid solution phase, our Xray diffraction analysis also suggests the formation of a Zn0.945Fe1.78O3.71 phase, which is an Fe-deficient ZnFe2O4 spinel phase. As FeO1.5 doping increases from 2 to 4 m/o, we observed a reduced fraction of Fe-deficient ZnFe2O4 spinel phase and a ZnO solid solution phase that has higher Fe concentration, suggesting a competition between the phase stabilities of ZnO and ZnFe2O4 spinel phases. The formation of Fe-deficient ZnFe2O4 spinel phase possibly occurs during cooling when passing through the ZnO/(ZnO + spinel) phase boundary, when some of Fe ions left oversaturated ZnO and formed Fe-deficient spinel phase that is energetically favored at lower temperature. Interesting phase change was found in the 18 m/o FeO1.5doped ZnO. When further annealed at 1350 °C and then cooled to room temperature, two ZnO solid solution phases with apparently different Fe concentrations were formed. The Fe highly concentrated ZnO, Zn0.750Fe0.171O, which has a much larger fraction than the less Fe concentrated ZnO solid solution phase, was not found in the sample annealed at 1150 °C nor in the 2 or 4 m/o FeO1.5 sample. This is attributed to the fact that equilibrated at 1350 °C, the 18 m/o FeO1.5 composition is nearly at the phase boundary reaching the large solubility limit 7932

DOI: 10.1021/am509050a ACS Appl. Mater. Interfaces 2015, 7, 7927−7937

Research Article

ACS Applied Materials & Interfaces

Fe concentration. The second step is cooling from 1350 °C to room temperature, during which a series of phase decomposition occurs, as seen from our X-ray diffraction and electron microscopy observations. As cooling starts, the major phase change is the precipitation of ZnFe2O4 spinel phase from the Fe highly enriched ZnO solid solution phase, obeying the (111) [110] // ZnO (0001)[112̅0] crystallographic relationship as observed from our electron microscopy studies. The EDS mapping of the 18 m/o FeO1.5 sample annealed at 1350 °C suggests that the volume fraction of spinel phase is ∼0.17, and the corresponding molar faction is 0.09; this molar fraction relates to ∼1300 °C on the phase diagram according to the Lever rule. This finding also suggests that there was formation of ZnO solid solution phase with ∼13 m/o FeO1.5 when cooled to this temperature. Nevertheless, a significant portion of Fe highly concentrated ZnO solid solution phase remained, which has ∼16.0 m/o FeO1.5 as discussed previously. As cooling continuously carries on, the Fe solubility in ZnO decreases and ZnO with less Fe content is favored by equilibrium. It turns out that a portion of ZnO solid solution phase changes to further lower Fe concentrated one. This is indicated by our X-ray diffraction result that the less Fe concentrated ZnO phase in the sample annealed at 1350 °C has only ∼0.6 m/o FeO1.5. The Fe ions leave the ZnO solid solution phase and form the Fe-deficient spinel phase, as is the case of 2 or 4 m/o FeO1.5 samples. Such a series of phase decomposition and change yields one of the most interesting phenomena reported in the present work: the microstructure in the 18 m/o FeO1.5 sample is significantly refined into nanoscale after high-temperature treatment (1350 °C), which is counterintuitive to the hightemperature coarsening phenomenon commonly observed in most ceramic materials. With samples cooled from 1350 °C, the ZnFe2O4 spinel phase either develops into the morphology of lamellar or platelet structures obeying the (111)[110] // ZnO (0001) [112̅0] crystallographic relationship or precipitates as nanograins with limited further growth due to the slowed diffusion process under cooling. Accordingly, original ZnO grains were then segmented into much smaller grains, and the average grain boundary spacing reduces by about an order of magnitude (from 2−3 μm to ∼200 nm). Thermal Conductivity. To investigate the compositional and structural roles in thermal transport properties, the roomtemperature thermal conductivity was plotted against the actual m/o FeO1.5 in ZnO lattices as shown in Figure 10b. Fe addition and annealing temperature of each sample are indicated beside each symbol. As seen from the figure, thermal conductivity generally decreases with the actual Fe content in ZnO lattices, which can be attributed to the point defects introduced into the ZnO lattices and the microstructural refinement (the formation of lamellar ZnFe2O4 spinel structure). Point defects, which result in an atomic scale change, are normally more influential in affecting phonon transport. In the current case of doping Fe2O3 into ZnO, the defect reaction is given by11

Figure 10. Effect of actual Fe content in ZnO lattices on the thermal conductivity. (a) ZnO basal plane (002) spacing or (c/2) measured by X-ray diffraction plotted as a function of actual m/o FeO1.5 in ZnO lattices. The Vegard law was determined based on the undoped ZnO and the 2 and 4 m/o FeO1.5-doped ZnO. The actual Fe content of ZnO solid solution in 18 m/o FeO1.5 sample annealed at 1150 and 1350 °C was then estimated. (b) Room-temperature thermal conductivity of Fe-doped ZnO versus actual m/o FeO1.5 in ZnO lattices. Each sample composition and annealing temperature is indicated beside the symbol. The solid green curve is the best fit of the measured data according to κ ∝ c−1/2.

ZnO (002) basal plane spacing or (c/2) of ZnO solid solution phases in the 18 m/o FeO1.5 sample onto the Vegard’s law, the actual Fe content in these solution phases can be estimated. The ZnO solid solution phase in the 18 m/o FeO1.5-doped ZnO annealed at 1150 °C has only ∼1.0 m/o FeO1.5 where much of Fe ions participate in the formation of ZnFe2O4 spinel phase. This can be attributed to the reducing Fe solubility with decreasing temperature. In the 18 m/o FeO1.5-doped ZnO annealed at 1350 °C, ∼0.6 m/o FeO1.5 content was estimated in the less Fe concentrated ZnO solution phase, which has a quite small fraction, whereas the Fe highly concentrated ZnO solid solution that is the major phase, is estimated to have ∼16.0 m/o FeO1.5. Microstructural Evolution of 18 m/o FeO1.5 Sample. The microstructural and phase evolution of the 18 m/o FeO1.5 sample annealed at 1150 and 1350 °C are relatively complicated but can be understood as a sequence of two major steps. The first step is the annealing at higher temperature, that is, at 1350 °C in the single-phase region. Clearly seen from the phase diagram in Figure 1a, as temperature increases from 1150 to 1350 °C, the two-phase mixture of ZnO solid solution phase and ZnFe2O4 spinel phase are transforming to a single ZnO solid solution phase with high

ZnO

Fe2O3 ⎯⎯⎯→ 2Fe•Zn + V″Zn + 3OOX + 3ZnO

(1)

which suggests doping one molecule of Fe2O3 into ZnO creates two substitutional iron point defects on zinc sites and one zinc vacancy for charge balance. The chemical formula of ZnO with Fe introduced point defects is then written as 7933

DOI: 10.1021/am509050a ACS Appl. Mater. Interfaces 2015, 7, 7927−7937

Research Article

ACS Applied Materials & Interfaces (Zn1 − 3c /2FecVc /2)O

(2)

A=

where c is the iron concentration. The point defect scattering strength factor Γ is characteristic of phonon scattering by point defects, and the general expression is given by32 Γ=



∑ fi ⎜⎝1 − i

Mi ⎞ ⎟ M⎠

(MFe − M Zn)2

(3)

2( M ZnO)2

+

2 c ( M Zn) 2 ( M ZnO)2

−S =

(4)

(5)

This suggests an increasing point defect phonon scattering with increasing iron concentration Zn lattices. In addition, the thermal conductivity κ is inversely proportional to Γ.33

κ ∝ Γ−1/2

(6)

This suggests that the thermal conductivity decreases with the Fe content c in the ZnO lattices, as seen by the best fit of the measured data, which is shown as a solid green curve in Figure 10b. Thermoelectric Properties. Relating the actual Fe solute concentration with the electrical conductivity data, we find that the electrical conductivity in general increases with the actual Fe content in ZnO lattices. For instance, the 18 m/o FeO1.5doped ZnO sample annealed at 1350 °C, in which the ZnO solid solution phase has the largest Fe content among all the samples studied, demonstrates the highest electrical conductivity. This is especially true as temperature goes up when there are an increasing number of electrons delocalized from the donor defect states and excited to the conduction band, leading to a larger free charge carrier concentration n. The role of ZnFe2O4 spinel phase is unclear, but it is not a very poor electrical conductor based on its measured band gap of ∼2.5 eV.34 A useful approach to look into the electrical transport process is to probe the relative position of the Fermi level in the band structure. Seebeck coefficient (also commonly known as thermopower or thermoelectric power in the literature), which is strongly related to the Fermi level in the materials,35,36 can be used to investigate the electrical conduction processes in both bulk35 and low-dimensional materials.37−39 The Seebeck coefficient for an n-type semiconductor expressed in terms of Fermi level EF as measured from the conduction band edge EC is given by36 S=−

⎤ kB ⎡ EC − E F + A C⎥ ⎢ e ⎣ kBT ⎦

E > EC (8)

EC − E F ⎛ 1 ⎞ ⎜ ⎟ + A ⎝T ⎠ e

(9)

where A = [((γg − γF/e) + AC(kB/e))]. The above equation suggests that the Fermi level as measured from the conduction band edge can be estimated from the plot of −S against 1/T. The Seebeck model in eq 9 assumes a single-phase material and therefore may be applied to ZnO, 2 m/o FeO1.5, and 4 m/o FeO1.5-doped ZnO in the present work. The modeling fit of the measured Seebeck coefficients as a function of temperature for these single-phase samples are presented in Figure 11a, where scattered symbols represent the measured data and dashed lines are modeling fit. The estimated Fermi levels are summarized in Table I. Our findings that the Fermi level is closer to the conduction band in the undoped sample can be attributed to native defects such as oxygen vacancies, which are commonly found in intrinsic ZnO.31,42 From 2 to 4 m/o FeO1.5 addition, we found a shift of Fermi level closer to the conduction band edge. This is consistent with the increased electrical conductivity because Fe ions are dissolved into ZnO lattices providing electron donors without forming the second phase. Since 18 m/o FeO1.5-doped ZnO samples are two-phase materials consisting of ZnO and ZnFe2O4 spinel phases, we need a model for describing the Seebeck behavior of a composite material on the dependence of each phase constituent. We can write the Seebeck coefficient of a twophase material based on Bergman and Fel’s model,43 which describes the electrical conductivity and thermoelectric power factor of two-phase composite:

where M ZnO is the mean mass of the doped ZnO formula unit and M Zn is the average atomic mass on zinc site, both of which are dependent on iron concentration c. The first-order approximation in math gives

Γ ≅ 0.67c

∫0 σ(E)dE

Considering the temperature dependence of bandgap and Fermi level, EC and EF in eq 7 can be replaced by (EC + γgT)40 and (EF + γFT),37,39 respectively, where γg and γF are temperature coefficients for the change of bandgap and Fermi level with temperature. (In fact, in the cases of large doping or carrier concentration, the temperature dependence of Fermi level becomes quite weak.41) Accordingly, eq 7 can be rewritten as

2

where Mi is the mass of the defect and M̅ is the average mass. In the current case, considering both the substitutional defects and zinc vacancy, Γ is written as Γ=c

∞ E − EC σ(E)dE kBT , ∞

∫0

S=

fA σASA + (1 − fA )σBSB fA σA + (1 − fA )σB

(10)

where σi and Si (i = A or B) represents the electrical conductivity and Seebeck coefficient of each phase, and fA is the volume fraction of the phase A. The difficulty in applying Bergman’s model is the unknown electrical conductivity of each phase, which can be significantly different from the intrinsic and undoped phase due to the electron doping in each phase. In practice, it is not a bad approach to apply the law of mixture in describing the Seebeck coefficient of a two-phase material; it has been reported that the law of mixture model on predicting Seebeck coefficient of a two-phase composite is a reasonably good estimation unless the electrical conductivity of the two phases differs by several orders of magnitude.44 Accordingly, it is not a bad approximation to apply this practical approach to our material system. The Seebeck coefficient of the FeO1.5 heavily doped ZnO, which consists of ZnO and ZnFe2O4 phases, is given by

(7)

where e is the electronic charge and kB is the Boltzmann’s constant. AC is a temperature-independent term named as the heat of transport constant, which is the contribution from the charge carriers that are distributed above the conduction band EC.

S = fspinel Sspinel + (1 − fspinel )SZnO 7934

(11)

DOI: 10.1021/am509050a ACS Appl. Mater. Interfaces 2015, 7, 7927−7937

Research Article

ACS Applied Materials & Interfaces

to the conduction band in the higher temperature annealed samples (1350 °C) as compared to the one annealed at 1150 °C. This is well consistent with the much higher Fe solute concentration observed in the higher temperature annealed sample (see Figure 10a) as well as its higher electrical conductivity (see Figure 9b). Also concluded from Table I, all the samples are nondegenerate semiconductors even with FeO1.5 addition up to 18 m/o, as Fermi levels all lie below the conduction band. We further perform the Jonker plot analysis through which the electrical conductivity is correlated with Seebeck coefficient.45−47 Electrical conductivity σ for semiconductor materials is given by the standard equation σ = neμ

(12)

where e is the electronic charge and μ is the electron mobility. Seebeck coefficient function is already presented in eq 7. To make it explicitly correlated with electrical conductivity, we shall re-express it in terms of carrier concentration n. In semiconductors, carrier concentration is dependent on the Fermi level through the well-known relationship below (for ntype semiconductor)48 ⎛ E − EF ⎞ n = NC exp⎜ − C ⎟ kBT ⎠ ⎝

where NC is the density of states at the conduction band edge given by Nc = (2πme*kBT/ℏ2)3/2 in which me* is the effective mass. Substituting eq 13 into eq 7, we can readily rewrite the Seebeck coefficient in terms of the carrier concentration.

Figure 11. Analysis of Seebeck coefficient of Fe-doped ZnO. (a) Plot of negative Seebeck coefficient (−S) against the inverse of absolute temperature 1/T. Measured data are shown by the scattered symbols of different colors; dashed lines of the corresponding colors are the modeling fit of the experimental data according to eq 9 for singlephase samples (ZnO and 2 m/o and 4 m/o FeO1.5-doped ZnO) and eq 11 for two-phase samples (18 m/o FeO1.5 samples). (b) Jonker plot (Seebeck coefficient against natural logarithm of electrical conductivity) of Fe-doped ZnO (including pure ZnO) measured at 800 °C. The best fit of the data with a slope of +86.15 μV/K is shown as the dashed line.

S=−

ZnO, 1150 °C 1 d 2 m/o FeO1.5, 1250 °C 1 d 4 m/o FeO1.5, 1250 °C 1 d 18 m/o FeO1.5, 1150 °C 1 d 18 m/o FeO1.5, 1350 °C 3 d a

(EC − EF)ZnO, eV

(EC − EF)ZnFe2O4, eV

0 0 0 0.18 0.17

0.066 0.133 0.092 0.196 0.100

0.538 0.100

⎤ kB ⎡ ⎛ Nc ⎞ ⎢ln⎜ ⎟ + A C⎥ ⎦ e ⎣ ⎝n⎠

(14)

Combining eqs 12 and 14, one immediately obtains the correlation equation between Seebeck coefficient and electrical conductivity. S=

kB (ln(σ ) − ln(σ0)) e

(15)

where σ0 = Nceμ exp(A) . This suggests a linear behavior when the Seebeck coefficient is plotted against the natural logarithm of electrical conductivity, which is commonly known as a Jonker plot.45 As shown in Figure 11b, all our data measured at 800 °C were made in a Jonker plot, and a reasonably good correlation between electrical conductivity and Seebeck coefficient is obtained with a linear slope of +86.15 μV/K, corresponding to the Seebeck value of classical free electron gas kB/e.49 This suggests that the electrical conduction process in these Fedoped ZnO materials are dominated by free electrons. The intercept, ln(σ0), referred as “DOS-μ” product, was found to have a value of 10.0, very similar to the value of ZnO−In2O3 system.50 In addition, Ioffe analysis46,47,51 suggests that the maximum achievable thermoelectric power factor PFmax is dependent on ln(σ0) given by the relation below.

Table I. Summary of Estimated Fermi Levels from the Temperature Dependent Seebeck Coefficient Measurementsa fspinel

(13)

The volume fractions of ZnFe2O4 spinel phase are also shown.

where Sspinel and SZnO are the Seebeck coefficients of ZnFe2O4 and ZnO phases, and fspinel is the volume fraction of the ZnFe2O4 phase. Sspinel and SZnO can be further expressed using eq 7. Similar to the single phase model shown in eq 9, we plot the −S of 18 m/o FeO1.5 samples as a function of 1/T, as shown by the scattered symbols in Figure 11a. The modeling fit of the measured data according to eq 11 are represented by the dashed lines, from which we estimated the Fermi levels in both ZnO and ZnFe2O4 phases, as also summarized in Table I. It is clearly found that the Fermi level in both phases is much closer

⎛ k ⎞2 ln PFmax = ln σ0 + ln 4⎜ B ⎟ − 2 ≈ −19.33 + ln σ0 ⎝e ⎠ (16) −4

This gives the PFmax ≈ 0.9 × 10 W/mK , which is a reasonably good value for oxide materials. 7935

2

DOI: 10.1021/am509050a ACS Appl. Mater. Interfaces 2015, 7, 7927−7937

Research Article

ACS Applied Materials & Interfaces The thermoelectric figure of merit ZT is calculated according to standard relation

ZT =

S2σ T κ

National Nanotechnology Infrastructure Network (NNIN), which is supported by the National Science Foundation under NSF Award No. ECS-0335765.



(17)

and plotted against temperature as presented in Figure 9d. The 18 m/o FeO1.5-doped ZnO sample annealed at 1350 °C has the best power factor and figure of merit significantly superior to other samples over the medium to high temperature range. Compared to other element-doped ZnO systems, it has relatively higher power factor or ZT value than Ga-doped ZnO52 and In and Ga dually doped ZnO,18 similar thermoelectric performance to Sb-15 or Sn16-doped ZnO, while lower than the Al-53 or In50-doped ZnO and Al and Ga dually doped ZnO.54 Nevertheless, the Fe−ZnO system has the great advantage of achieving the refined nanostructures through high temperature treatment, which is essential to maintain good thermoelectric properties at high temperatures.

(1) Tsubota, T.; Ohtaki, M.; Eguchi, K.; Arai, H. Thermoelectric Properties of Al-doped ZnO as a Promising Oxide Material for Hightemperature Thermoelectric Conversion. J. Mater. Chem. 1997, 7, 85− 90. (2) Fujishiro, Y.; Miyata, M.; Awano, M.; Maeda, K. Effect of Microstructural Control on Thermoelectric Properties of Hot-Pressed Aluminum-Doped Zinc Oxide. J. Am. Ceram. Soc. 2003, 86, 2063− 2066. (3) Park, K.; Choi, J. W.; Kim, S. J.; Kim, G. H.; Cho, Y. S. Zn1‑xBixO (0 < x